Organize a workspace for repeatable AI tasks

A simple folder structure makes source material and generated work easier to distinguish. Keep inputs, drafts and reviewed deliverables in separate locations.

Start with a small handoff

Select a role by its working guide and deliverables, then verify the connected AI tool and workspace. Keep the first run small enough that you can inspect every output. The desktop needs to stay running while work executes. Marketplace sign-in, model access and file-generation tools are separate setup requirements; a success in one does not prove the others are ready.

Inputs for this workflow: A dedicated project folder and current source files.

Work through the task

1. Create clear input and output locations.
2. Define how drafts become reviewed deliverables.

A brief you can adapt

Read files under inputs and save proposed results under outputs/drafts. Write a manifest of generated files and do not replace reviewed deliverables.

The filenames above are examples. Replace them with your actual inputs and destination before submitting the task.

Review the deliverable

Check the manifest against the actual folder and confirm that source files were preserved.

Where this approach can fail

Folder naming is a workflow convention, not a security boundary for a locally running CLI.

For the related desktop setup, see first task.